This article appeared in a journal published by Elsevier.

The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elseviers archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy

Journal of Asian Economics 23 (2012) 395408

Contents lists available at SciVerse ScienceDirect

Journal of Asian Economics

Financial reforms and persistently high bank interest spreads in Bangladesh: Pitfalls in institutional development?
Monzur Hossain *
Bangladesh Institute of Development Studies (BIDS), E-17 Agargaon, Sher-e-Bangla Nagar, Dhaka 1207, Bangladesh

A R T I C L E I N F O

A B S T R A C T

Article history: Received 3 December 2010 Received in revised form 28 November 2011 Accepted 29 December 2011 Available online 6 January 2012 JEL classication: G21 G30 O16 Keywords: Interest rate spread and margin Financial liberalization Bank efciency Bangladesh

This paper analyzes interest rate spreads and margins in Bangladesh for the period 1990 2008 by applying the ArellanoBover/BlundellBond dynamic panel regression model to a panel of 43 banks. The model has been applied to tackle short-panel bias and endogeneity problems in banking analysis. A high degree of persistency in spreads and margins is observed, which points to inefciencies of bank management. More specically, high administrative costs, high non-performing loan ratio, market power, small share of deposits and some macroeconomic factors are found to be the key determinants of persistently high interest rate spreads and margins in Bangladesh. The ndings of this study suggest that reforms commenced in the 1990s could not generate adequate competition and efciency in the nancial sector, particularly to drive down the spread in line with the predictions of interest rate literature. This situation in other words indicates pitfalls in institutional development. 2012 Elsevier Inc. All rights reserved.

1. Introduction It is widely argued that nancial reforms and liberalization that involve decontrolling interest rates, eliminating credit limits and enacting new laws and regulations governing the nancial sector should improve efciency in the intermediation process. The interest rate spread (IRS), i.e., the difference between the weighted average interest rates on loans and deposits, is a key indicator of nancial performance and efciency of the banking sector. The spread is expected to decline over time with liberalization of the nancial sector. This proposition is linked to the McKinnon (1973)Shaw (1973) paradigm that nancial liberalization leads to signicant improvement of growth prospects. A high spread usually refers to a low deposit rate and a high lending rate that act as impediments to the expansion of nancial intermediation by entailing a high cost of borrowing and discouraging savings in the economy. A high spread thus limits investment opportunities and restricts the growth potential of the economy. The conventional view is that nancial liberalization and growth usually go together as liberalization increases the supply of loanable funds to the economy through increasing efciency of the nancial sector (Khan & Senhadji, 2000; King & Levine, 1993; Levine, 1997). However, nancial liberalization may not lead to the expected outcome unless necessary legal and nancial institutions are properly developed (Chinn and Ito, 2006). Bangladesh carried out extensive nancial sector reforms during the 1990s. Financial liberalization measures adopted include lifting barriers to entry of foreign and private commercial banks, decontrolling interest rates and credit disbursement, unication of exchange rates and adopting more exible exchange rates and making the current account

* Tel.: +880 2 8129625. E-mail addresses: monzur@bids.org.bd, monzur71@gmail.com. 1049-0078/$ see front matter 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.asieco.2011.12.002

Author's personal copy

396 M. Hossain / Journal of Asian Economics 23 (2012) 395408

convertible. However, the extent of interest rate spread has not changed much in Bangladesh despite these measures.1 The average interest rate spread was estimated to be 6.13 percent in the 1980s, 6.37 percent in the 1990s and 5.35 percent in the 2000s.2 From the concern that a large interest rate spread is an impediment to growth prospects, Bangladesh Bank (the central bank of Bangladesh) had been persuading banks to reduce the spread in a rational manner. As moral suasion did not work, Bangladesh Bank imposed ceilings on both lending and deposit interest rates in 2008; however, the ceilings were removed in 2011 mainly due to pressure from the IMF. This back-and-forth strategy towards interest rates raises several questions: why are spreads and margins persistently high in the banking sector of Bangladesh despite nancial liberalization? Why do different types of banks charge different interest spreads? A wide range of studies identied that large spreads occur in developing countries mainly due to high operating costs, nancial taxation or repression, lack of a competitive nancial/banking sector and macroeconomic instability (Barajas, Steiner, & Salazar, 1999; Beck & Hesse, 2009; Brock & Rojas-Suarez, 2000; Chirwa & Mlachila, 2004). However, our understanding of the determinants of spreads and margins in Bangladesh is still limited. There is no comprehensive, rigorously conducted analysis of spreads and margins currently available for Bangladesh in part because of data limitations.3 Two recent studies, Mujeri and Islam (2008) and Mujeri and Younus (2009) shed some light on the characteristics of interest spreads in Bangladesh. This paper makes an attempt to improve our understanding of the spreads and margins in Bangladesh by analyzing a unique panel data set relating to 43 banks for the period 19902008.4 A generalized method of moments (GMM) dynamic panel regression model, namely the Arellano and Bover (1995)/Blundell and Bond (1998) model, has been applied to the data to identify the determinants of spreads and margins as well as to capture their persistency. Data have been collected from the commercial banks balance sheets and income statements. This study, for the rst time, captures the persistency of spreads and margins in the banking sector of Bangladesh by applying dynamic GMM estimators. The estimated persistency effect (0.42) indicates that a major part of interest spreads in Bangladesh can be explained by some unobserved characteristics of the banking sector including inefciencies of management arising from revealed preferences, weaknesses in risk management practices and technological skills. In general, both less-competitive market structure and management inefciency are held responsible for persistently high spreads in Bangladesh. As a result of nancial liberalization, market power has shifted from the state-owned commercial banks (SCBs) to the old but big private commercial banks (PCBs) in the post-liberalization period (after 1999). This indicates that nancial reform measures undertaken in the 1990s have not contributed much to make the sector more competitive. The rest of the paper is organized as follows. Section 2 provides a brief survey of literature on interest rate spreads. Section 3 provides an overview of the nancial sector reforms and development in Bangladesh. Section 4 discusses data and variables and Section 5 discusses methodology and results. Finally, Section 6 concludes the paper. 2. A brief survey of literature What are the determinants of spreads and margins? Does nancial liberalization decrease the level of spread? These two questions are addressed in most studies dealing with interest spreads. A review of the determinants of spreads is provided in Table A.1 in Appendix. Beck and Hesse (2009) categorize the determinants of spreads and margins under four broad-based views. First, the riskbased view captures some systematic differences across borrowing sectors and deciencies in the contractual and informational frameworks driving high spreads and margins. According to this view, bank size, capital ratio, bank liquidity, operating costs, non-performing loan (NPL) and non-interest income are associated with risk management practices of banks. Second, the small nancial system view focuses on the xed transaction cost component of nancial service provision and the difculties in exploiting the resulting scale economies. The market share of deposits and/or loans usually represents the size of the nancial system. Third, the market structure view usually focuses on the competitiveness and the extent of privatization and foreign bank entry into the banking system. Market concentration ratios are used to assess the relevance of this view towards spreads and margins. Finally, the macroeconomic factors such as exchange rates, interest rates, ination rates and GDP growth rates are also considered as driving forces of interest spreads and margins in the banking system. All these factors together or partially can contribute to high spreads and margins in a less developed nancial system. Are the determinants similar across countries? Interest spreads are higher in developing countries than developed countries. Among developing countries, spreads are higher in African and Latin American countries than those in Asian countries. It can be observed from Table A.1 in

1 Banks were allowed to adjust their own rates since February 19, 1997. Further exibility in the interest rate was introduced on July 12, 1999 permitting banks to differentiate interest rates to individual borrowers except exporters (Economic Trends, Bangladesh Bank). 2 Interest rate spreads in Bangladesh are comparable to other South Asian countries. The average spreads for the last ve years was 6.0 percent in Pakistan, 4.95 percent in India and 6.18 percent in Sri Lanka (source: respective central bank). Thus the Bangladesh case is nothing but a typical South Asian case of maintaining moderate but persistent level of spreads. 3 In recent days, a growing tendency can be seen among banks to be engaged in capital market businesses through operating merchant banks, creating mutual funds and trading individually in the stock markets. However, their prots from share-market business are not clearly reported in any of the published documents. 4 A total of 48 banks are now operating in Bangladesh, of which long time-series data are available for 43 banks.

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408 397

Appendix that almost similar factors such as management inefciency, high administrative costs, high non-performing loans, market power and ination can explain high spreads across countries. Country-specic characteristics do not seem to have important implications for higher interest rate spreads. Studying the determinants of spreads and margins is meaningful only in a nancially liberalized economy. The empirical evidence regarding the impact of nancial liberalization on spread is mixed. While some studies argue that nancial liberalization substantially reduces spreads (see, Denizer, 1999; Honohan, 1999), some other studies reported the opposite scenario (e.g. Barajas et al., 1999; Chirwa & Mlachila, 2004). The contrasting evidence can be explained by the difference in the level of nancial reforms, regulatory framework in place, institutional strength and country-specic factors. Regarding econometric techniques, it is observed that various types of regression models are used to analyze interest spreads. Some of the common methods are pooled OLS (Ordinary Least Squares) regression, median least squares method, xed effect (FE) and random effect (RE) panel regressions and system equation (Table A.1). These regression models largely suffer from short-panel bias and they are also not suited to tackle endogeneity problem. The interest rate spread is associated with both observed and unobserved characteristics of banks including risk aversion attitude and revealed preferences of managers, governance structure etc., which cannot be captured unless a suitable model is applied. Consider some examples of endogeneity in banking variables. The capital structure of banks acts as a buffer against failure of banks. Non-performing loan is an ex post measurement of the risk assumed by the institution. These variables are likely to be correlated with idiosyncratic component of risk prole of institutions, hence, they are endogenous. Overhead costs, bank size, non-interest income also appear to be endogenous as they are the choice variables. Moreover, persistency of spreads is harmful to the economy even though the spread is moderate. Persistency in spreads refers to unobserved characteristics of banks, such as managerial risk aversion and revealed preferences. Therefore, it is important to capture the persistency effect of spread. If a method is used without taking into account the concerns raised here, it may lead to biased and inefcient estimates. One of the solutions to address the persistency and endogeneity issues could be the use of the GMM dynamic panel model. 3. Financial reforms, nancial development and interest spread in Bangladesh The formal nancial sector in Bangladesh, as in other regions of the developing world, essentially consists of banks. Although non-bank nancial institutions and capital markets have been developing gradually in Bangladesh, their inuence in the economy still remains marginal compared to the banking sector. The banking sector at present comprises of 48 banks including 4 state-owned commercial banks (SCBs), 30 private commercial banks (PCBs), 5 specialized banks (SBs) and 9 foreign commercial banks (FCBs) (see details in Table 1). Private banks were allowed to operate in Bangladesh from the early 1980s.
Table 1 Characteristics of the nancial sector of Bangladesh. Bank type Number Number of branches Rural A: Financial intermediation in Bangladesh (as of March, 2009) State owned commercial banks 4 2146 Private commercial banks 30 634 Specialized banks 5 1206 Foreign commercial banks 9 0 Total Period average 48 Credit to private sector (% of GDP) 3986 Urban 1240 1461 157 56 2914 Total 3386 2095 1363 56 6900 30.66 53.71 6.08 9.55 100.00 Gross xed capital formation (%GDP) 10.44 10.51 13.87 17.93 21.51 22.63 24.4 Interest rate on exports 7 7 7 7 7 Interest rate on trade nancing 12.49 12.59 14.30 14.41 14.07 48.07 29.71 8.31 13.91 100.00 GDP per capita at current US dollar 160.0 192.0 242.0 283.0 353.0 395.0 565.5 Interest rate on house nancing 10.02 10.15 12.95 12.98 12.85 Interest rate on consumers loan 7.29 8.81 13.66 14.16 14.56 Percentage of total asset Percentage of total deposit

Total deposits (% of GDP) 14.86 20.23 24.75 23.07 26.7 35.08 45.0 Interest rate on large term loan Interest rate on small term loan

Broad money (% of GDP) 19.03 24.54 28.67 26.68 31.01 40.02 45.0 Interest rate on working capital 11.88 12.01 13.59 13.75 13.07

B: Financial development in Bangladesh 19761980 6.59 19811985 13.67 19861990 19.08 19911995 16.58 19962000 23.17 20012005 28.83 20062008 34.5 Year Savings rate Fixed (term) deposit rate Interest rate on agri. loan

C. Interest rate structure across banks in Bangladesh (yearly average) 2004 5.50 7.60 9.37 11.50 10.88 2005 5.56 7.91 9.41 11.61 10.97 2006 5.99 9.59 9.92 13.19 12.08 2007 5.99 9.82 9.93 12.90 11.98 2008 5.95 10.98 10.41 12.48 12.10

Sources: (1) Economic trends, Bangladesh Bank (various issues); Bangladesh Bank Bulletin (various issues), (2) Bangladesh economic review (various issues), Ministry of nance.

Author's personal copy

398 M. Hossain / Journal of Asian Economics 23 (2012) 395408

Fig. 1. Non-interest income of different banks. Notes: SCB indicates the state-owned commercial banks, PCBs indicate private commercial banks and FCBs indicate foreign commercial banks. Non-interest income represents the ratio of commission and fees as percentage of interest income. Source: Authors calculation.

Fig. 2. Bank interest spreads and margins in Bangladesh. Note: data of 43 commercial banks for the period 19912008 are considered. SCBs indicate the state-owned commercial banks, PCBs indicate private commercial banks, FCBs indicate foreign commercial banks and SBs indicate specialized banks. For denitions of interest spread and margin, see notes in Tables 3 and 4. Source: Authors calculation. (A) Bank interest rate spread in Bangladesh. (B) Bank interest margin in Bangladesh.

Prior to reforms that began in the early 1990s, banks were mostly government-controlled and political imperatives were consistently given priority over commercial viability. Competition between banking institutions remained stied and banks had little incentive to develop their activities. As a result, the institutional capacity of banks to manage the systemic and idiosyncratic risks in nancial systems has failed to develop sufciently. In part to remedy these problems, Bangladesh pursued extensive nancial sector reforms in the 1990s. These reforms generally entailed nancial liberalization, institutional reforms and prudential regulatory frameworks.5 The main features of these reforms were interest rate deregulation, relaxation of regulations on credit disbursements, strengthening of capital base of banks, adoption of risk management guidelines and initiation of exible exchange rates. These have succeeded in limiting the scope of government intervention in the nancial sector and in strengthening prudential regulation of nancial institutions. As a result of liberalization, the dominance of SCBs has reduced after 1999 with a strong emerging role of PCBs (see Table 1). Interest rate deregulations were done in steps. Initially the banks were allowed to set lending and deposit interest rates within certain bands. Later the bands were removed allowing the banks to determine interest rates along the lines dictated by market conditions. Financial liberalization process was completed in 1999 by removing all other restrictions that enable the banks to enjoy greater exibility in determining interest rates. Financial development indicators display steady increasing trend, implying widening and deepening of the nancial system in Bangladesh over time (Table 1, Panel B). It is observed that the average credit, deposit and broad money to GDP ratios increased substantially from 6.6 percent, 14.9 percent and 19.0 percent to 28.8 percent, 35.01 percent and 40.0 percent, respectively during 19762005. Investment as a percentage of GDP and per capita income (in current USD) also display a similar pattern and move broadly together reecting a close association among nancial development, investment and per capita income during the period.

5 These reforms were done under the Financial Sector Reform Program (FSRP) in the 1990s. This is a World Bank led reform program within the context of structural adjustment program.

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408 399

Interest rates (yearly average) for different types of credits and deposits are reported in Table 1 (Panel C) for the period 20042008. It can be observed that interest rates on trade nancing, working capital and consumer loan are higher than those on other types of loans and advances. On the other hand, savings rates remained fairly stable ranging between 5 and 6 percent during the period considered, while the xed (term) deposit rates showed an increasing trend. Non-interest income (e.g. commission and fees) is substantially higher among private commercial banks, particularly among FCBs (Fig. 1). The estimated spreads and margins showed slightly decreasing trend particularly after 1999 (Fig. 2). After 2004, spread was estimated to be less than 5 percent in the case of PCBs, but over 5 percent in the case of other banks. The highest spread is observed for FCBs (8.83%).6 Interest margin also showed an increasing trend except for 2007. It is apparent from Fig. 2 that FCBs enjoy higher margins than their counterparts. The above discussion suggests that although nancial liberalization and reforms have improved nancial deepening and diversication in Bangladesh, nancial market still remains segmented and less competitive. 4. Data and variables The data on 43 banks for the period between 1990 and 2008 are used for the analysis.7 Interest rate spreads and margins are our dependent variables. The spread is dened as the difference between the weighted average lending rate and weighted average deposit rate, where the weights are the relative amounts of loans or deposits contracted at specic interest rates in the respective year by the respective bank. On the other hand, the net interest margin is dened as the difference between total interest income plus commission/fees received over total earning assets and total interest paid minus commission/fees over total interest bearing liabilities. Explanatory variables include variables representing risk factor, market structure, small nancial system and macroeconomic conditions, as suggested by the literature. The variables that represent risk-prole of banks are, inter alia, the bank size, capital ratio, bank liquidity, operating costs, non-performing loans (NPL) and non-interest income. The logarithm of total asset is used as a measure of bank size. Given the operating efciency of banks, bank size can inuence the spreads either negatively or positively depending on the scale of economies. If scale economies work, bigger banks can maintain lower spread, otherwise not. Similar contrasting results can be observed in the case of non-interest incomethe ratio of commission and fees to interest income. A bank which has higher non-interest income may not be keen to earn interest income especially if there is market segmentation and inadequate competition. In fact, non-interest income may lead to either a high spread or a low spread depending on market conditions. Hence, the signs of coefcients for these variables may be either positive or negative. Overhead cost is the ratio of administrative costs to total assets. Higher operating costs are expected to lead banks to charge higher interest spreads. High overhead cost may result from inefciencies in bank operations that may be shifted to bank customers. Liquidity and capital ratio are indicators of bank solvency. Bank liquidity is dened as the ratio of total operational assets to total bank liabilities. This variable is expected to be negatively related to interest spread. An increase in liquidity reduces the bank liquidity risk, which reduces the interest spread due to a lower liquidity premium charged on loans. Capital ratio is dened as the ratio of shareholders equity to total assets. Saundars and Schumacher (2000) provide evidence of the positive and generally signicant relationship between spreads and capital ratios in developed countries. Since there are limited channels for increasing capital because of thin and underdeveloped equity markets in developing countries, banks will be in a strong position to keep the spreads high. Thus, the capital ratio is expected to be negatively associated with the spread. As liquidity appears to be highly correlated with capital ratio, only capital ratio is considered in the analysis. Historically, banking sector in Bangladesh is characterized by high non-performing loans, most of which are borne by the state-owned commercial banks (Fig. 3). While NPL ratio is about 5 percent for PCBs, it is still about 20 percent for SCBs (Table 2). Banks tend to offset the cost of screening and monitoring of bad loans and/or the cost of foregone interest revenue by charging higher lending rates (Barajas et al., 1999). Therefore, NPL might have a positive and signicant association with spreads (Brock & Rojas-Suarez, 2000; Randall, 1998). The market share of deposits or loans is usually used to see whether the size of the nancial system matters for the spread. The market share of loan (deposit) is the share of individual banks loans (deposits) to total loans (deposits) provided by both banks and non-banks in a year. The market share of banks in total deposits is estimated to be about 55 percent on average during the time period considered, and the rest of the deposits are attributed to public borrowings through NSD certicates8 and postal deposits. Among the banks, while SCBs share in total deposits is 29 percent, PCBs share is 21 percent (Table 2). Historically, four SCBs capture the major share of both deposits and loans in Bangladesh. In the analysis we consider only the market share of deposits (MSD), which is estimated as a ratio of individual banks deposits to all deposits including deposits in

6 It was not possible for us to estimate the weighted average spread for the FCBs due to unavailability of their data on loans and deposits. Some of the FCBs do not even publish country-specic annual reports; they only prepare performance reports for the Bangladesh Bank. 7 Data are not available for all banks for all the years as some new banks have emerged during the period considered. Moreover, all the required data are not available for all banks particularly for the period before 1999. Thus, the panel is unbalanced. 8 The National Savings Directorate (NSD) certicates are the principal devices of public (non-bank) borrowing for nancing budget decit. The interest rate on 3-year NSD certicate has been 11.5% while the same on 5-year certicate is 12%. These savings rates are substantially higher than those are offered by banks (see Table 1).

Author's personal copy

400 M. Hossain / Journal of Asian Economics 23 (2012) 395408

Fig. 3. Non-performing loan ratio for different banks.Note: data of 43 commercial banks for the period 19932008 are considered. SCBs indicate the stateowned commercial banks and PCBs indicate private commercial banks. Non-performing Loan (NPL) represents the ratio of bad loans to total loans. Source: Authors calculation.

banks and non-banks, postal deposits and public borrowings through National Savings Directorate (NSD) certicates. The variable MSD is particularly important for capturing the impact of NSD certicates on spreads, as researchers and practitioners often argue that by offering higher than market interest rates for NSD certicates, the government creates distortions in the money market that ultimately inuence the spread. The NSD certicate rates, which are higher than bank deposit rates, are likely to lter deposits away from banks. This may inuence the spread by causing a rise in bank lending rates due to limited supply of loanable funds. While a negative relationship between market share and interest rate spreads predicts a small nancial system, a positive relationship would predict a less competitive market structure. To assess the impact of market structure on spread, market concentration ratios for loans and deposits are estimated considering the Herndahl-Hirschman Index (HHI). Both indexes indicate that market power has shifted gradually from SCBs to PCBs after nancial liberalization (Fig. 4). Particularly, PCBs concentration ratios for loans and deposits have crossed the one for SCBs in 2004. Since the HHIs for PCBs hovered around 4000 in 2008 with an increasing trend, it indicates that a monopolistic competition prevails in the banking sector of Bangladesh. In addition, the impact of nancial liberalization is captured by a nancial liberalization dummy (FLI) (1 for year 1999 onward, 0 otherwise). Among macroeconomic factors, quantum index of production (QI), ination, liquidity reserve requirement (LRR) and corporate income tax rate (Tax) are considered as potential drivers of spreads and margins. The GDP growth rate has not been used in the analysis as it may not reect the impact of overall demand of the economy on spreads due to segmented credit
Table 2 Summary statistics of key variables. Stats Interest spread Overhead cost as % of total asset Capital ratio as % of total asset 0.20 1.30 6.39 0.14 0.12 0.86 0.04 0.03 0.89 0.22 1.57 7.05 0.42 3.08 7.35 0.16 0.16 1.01 0.21 1.92 9.33 NPL as % of total loan 9.00 0.10 1.23 16 0.12 0.75 21 0.11 0.55 7 0.10 1.36 3 0.03 0.85 8 0.10 1.31 7 0.10 1.38 Liquidity reserve ratio (%) 19.57 1.73 0.09 22.21 0.60 0.03 22.45 0.61 0.03 20.01 1.13 0.06 19.48 1.35 0.07 20.12 1.07 0.05 20.25 1.09 0.05 Non-interest income as % of interest income 26 0.17 0.67 23 0.12 0.50 25 0.09 0.34 24 0.13 0.55 20 0.11 0.54 25 0.14 0.55 21 0.09 0.43 MSD

All banks Mean 5.18 0.05 sd 1.70 0.48 cv 0.33 9.08 State owned commercial banks Mean 5.83 4.00 sd 0.91 0.02 cv 0.16 0.66 State-owned commercial banks (year>1999) Mean 5.45 2.00 sd 0.81 0.01 cv 0.15 0.34 Private commercial banks Mean 5.01 6.00 sd 1.82 0.58 cv 0.36 9.18 Private commercial banks (New) Mean 3.75 14.0 sd 2.00 1.14 cv 0.53 8.34 Private commercial banks (Old) Mean 5.42 4.0 sd 1.56 0.02 cv 0.29 0.62 Private commercial banks (Year > 1999) Mean 4.62 7.0 sd 1.88 0.71 cv 0.41 10.40

0.54 0.23 0.42 0.29 0.16 0.54 0.25 0.15 0.54 0.21 0.10 0.48

0.22 0.11 0.51

Notes: sd, standard deviation; cv, coefcient of variation. See notes in Table 3 for denitions of variables. Source: Authors estimation.

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408 401

Fig. 4. Herndahl-Hirschman (HH) index for loans and deposits. Note: data of 43 commercial banks for the period 19912008 are considered. SCBs indicate the state-owned commercial banks, PCBs indicate private commercial banks, FCBs indicate foreign commercial banks and SB indicates specialized banks. To understand the market structure, the market concentration ratios are estimated by the Herndahl-Hirschman Index (HHI) for deposits and loans. Source: Authors calculation. (A) HH Index for loans. (B) HH Index for deposits.

markets with subsidized interest rates (see Table 1). Instead, the QI is used to capture the impact of segmented credit market on spreads as banks charge a competitive interest rate for industrial credit. A positive association of QI with spread will indicate the existence of segmented credit market. Correlation between variables is estimated in Table 3. Table 3 (Panel A) shows that before 1999, the spreads were correlated positively with the loan rate and negatively with the deposit rate (except for SBs). After 1999, in the case of PCBs, spread is found to be positively correlated with the rates of large loans and working capital loans but negatively with savings deposit rates. In the case of SCBs, spread is correlated (negatively) only with savings deposit rates. In the case of SBs, spread is perfectly correlated with large loan rate. A high interest spread is therefore associated with the rise of the lending rates of large loans or decrease of the deposit rates. Table 3 (Panel B) shows pair-wise correlation between the variables concerned. In most cases, the correlations between spread and other variables show a positive and signicant relationship, but far from perfect correlation.

5. Methodology and results 5.1. Methodology One of the difculties that one may have to encounter in the banking sector analysis is that all unobserved bank characteristics are not captured in the available data. Another difculty is that most of the banking variables, such as interest spread, prot margin, capital ratio, NPL etc. are endogenous, that is, these variables are likely to be correlated with unobserved rm-level heterogeneity. If the OLS method is used ignoring such unobserved rm-level characteristics, it will lead to biased and inconsistent estimates (Wooldridge, 2002). The application of the xed effect models to a dynamic panel is also problematic. One potential problem of the xed effect model is that rm xed effect is correlated with the lagged dependent variable, which introduces a bias that is substantial with shorter panels9 (Baltagi, 2005; Nickell, 1981; Wooldridge, 2002). Since in our data each panel consists of about 40 observations, we need to apply a suitable model that can address the biases in estimates due to a short panel as well as endogeneity problems. There are a number of choices: The rst option is to adopt a traditional instrumental variables (IV) approach. However, in corporate nance, it is difcult to nd reliable instruments. The other option is to apply the generalized method of moments (GMM, or differenced GMM) estimator suggested by Arellano and Bond (1991). This GMM technique transforms the data into rst-difference form and then uses the endogenous (or predetermined) lagged variables as instruments for the transformed lagged dependent variable. However, the ArrelanoBond model also suffers from certain limitations. The lagged levels provide little information about the rst differences when the underlying series are relatively stationary and, therefore, are weak instruments (Arellano & Bover, 1995; Blundell & Bond, 1998). To overcome the problem particularly in a shorter panel, Arellano and Bover (1995)/Blundell Bond (1998) modied the GMM by employing additional moment conditions based on the lagged variables rst differences (in addition to their levels). Thus, to address the short-panel bias and endogeneity problem, here we choose to apply the ArellanoBover/Blundell Bond GMM model. The regression model is thus specied as:

IRSi;t a bBi;t g It dM t ei;t

This bias can be quite large even for panels with 30 observations per unit (Judson and Owen, 1999).

Author's personal copy

402 Table 3 Correlations of variables. Loan rate Agriculture Large loans Small loans Working capital Savings Deposit rate Fixed term M. Hossain / Journal of Asian Economics 23 (2012) 395408

A. Correlation of spread with lending and deposit rates Panel A. Before 1999 Private com. banks 0.50 0.38 State-owned com. banks 0.05 0.47 Specialized banks 0.39 0.99 Overall Panel B. After 1999 Private comm. banks State-owned com. banks Specialized banks Overall Interest spread B. Pair-wise correlation Overhead Capital ratio NPL Bank size Non-interest income MSD HHI QI Ination LRR Tax 0.10 0.25 0.03 0.39 0.17 Over head Capital ratio 0.04 0.39 0.14 0.99 0.90 NPL

0.54 0.52 0.61 0.40 0.05 0.10 0.61 0.07 Bank size

0.44 0.43 0.56 0.27 0.48 0.19 0.56 0.22 Non interest income MSD HHI

0.45 0.57 0.47 0.47 0.51 0.44 0.47 0.24 QI

0.53 0.61 0.59 0.43 0.05 0.34 0.59 0.02 Ination LRR

between bank-specic variables 0.12* 0.11* 0.99* 0.16* 0.15* 0.12* * 0.05 0.21 0.24* 0.24* 0.001 0.03 0.15* 0.01 0.02 0.16* 0.01 0.002 0.08 0.01 0.01 0.23* 0.01 0.01 0.06 0.003 0.03 0.12* 0.01 0.002

0.19* 0.01 0.30* 0.21* 0.05 0.05 0.20* 0.03

0.15* 0.57* 0.45* 0.06 0.23* 0.88* 0.02

0.01 0.05 0.03 0.001 0.08 0.13*

0.86* 0.14* 0.03 0.56* 0.03

0.20* 0.09* 0.44* 0.02

0.15* 0.05 0.15*

0.25* 0.23*

0.05

Notes. Data of 43 commercial banks for the period 19902008 are considered. The interest spread is calculated by taking difference between the weighted average loan rate and weighted average deposit rate for each bank and each year, where the weights are the relative amounts of loans or deposits contracted at specic interest rates in the respective year and by the respective bank. The logarithm of total asset is used as a measure of bank size. Non-interest income implies the ratio of commission, fees over interest income. Overhead cost is the ratio of administrative costs to total assets. Non-performing Loan (NPL) ratio represents the ratio of bad loans over total loans. Capital ratio is dened as the ratio of shareholders equity to total assets. The market share of deposits (MSD) is the share of individual banks deposit in a year in terms of total deposits including deposits in banks, non-banks, postal deposits and National Savings Directorate certicates. To understand the market structure, the market concentration ratios are estimated by the Herndahl-Hirschman Index (HHI) for deposits and loans. QI is the quantum index of industrial production. LRR is the liquidity reserve requirements, set by the Central Bank. Tax indicates corporate income tax for banks. Source: Authors estimation. * Represent signicance at 10% level. ** Represent signicance at 5% level. *** Represent signicance at 1% level.

where Bi,t is a vector of bank-specic variables, such as overhead costs, liquidity ratio, capital ratio, bank size, NPL, MSD, and non-interest income; It is a vector of time-varying market structure variables, such as FLI and HHI; Mt is a vector of timevarying macroeconomic variables, such as QI, GDP growth rate, ination, corporate tax rate and LRR. The QI is used to see whether industrial production has any impact on interest spread. The ArellanoBover/BlundellBond GMM model provided tests for AR(1) and AR(2) in rst differences. The model introduces rst order serial correlation; however, the test for no second order serial correlation for the disturbances of the rst-differenced equations is important for testing the consistency of the GMM estimates. The results show that there exists rst order serial correlation, but not the second order serial correlation. Further, the Sargan (1958) test has been applied to test the joint validity of moment conditions (the presence of over-identication) and to identify optimal lag. The tests conrm that the instruments used are orthogonal to the error term, that is, over-identication is rejected. The optimal lag is found to be two years in most of the cases. 5.2. The results The results are obtained by analyzing spreads separately for different categories of banks, and for separate time periods period before 1999 (pre-liberalization period) and after 1999 (post-liberalization period) to capture the behavior of different banks in the pre- and post-liberalization periods. Although nancial liberalization was completed in 1999, different liberalization measures were taken in steps before 1999. Therefore, the results representing the period 19901999 are attributed to partial liberalization of the nancial sector. Table 4 reports the results for all banks. The results provide support to all the underlying views of the determinants of interest spreads and margins in Bangladesh. The effects of lagged interest spreads and margins are found to be signicant, indicating persistency in interest rate spreads and margins. The estimated persistency effects on spreads and margins are, 0.42 and 0.63, indicating that last years spread and margin will amplify current spread and margin by 42

Table 4 Determinants of interest spreads and margins for all banks. Interest margin All banks (19902008) 0.41 (0.06)*** 10.30 (7.04) 1.04 (1.11) 2.32 (1.26)** 0.17 (0.75) 0.12 (0.90) 1.21 (0.31)*** 0.001 (0.001) 0.04 (0.12) 0.02 (0.03) 0.49 (0.32)* 0.42 (0.25) 0.02 (0.03) 3.97 (2.59) 12.13 (4.04)*** 0.97 (0.35)*** 0.03(0.04) 3.66(5.04) 7.28 (4.88) 220 123.7*** 140.85*** 332 2127.41*** 357.82*** 300 158.08*** 193.01*** 80 48.65*** 52.55*** 332 2203.4*** 357.17*** 0.33 (0.09)*** 3.98 (6.59) 1.90 (0.73)*** 1.80 (1.21) 2.39 (1.01)*** 0.36 (0.69) 0.28 (0.52) 0.01 (0.001)** 0.01 (0.01)*** 0.07 (0.02)*** 0.08 (0.23) 0.11 (0.27) 1.34 (0.64)*** 5.11 (7.89) 19.76 (6.30)*** 0.36 (0.08)*** 20.81 (9.12)*** 1.36 (1.87) 3.00 (1.60)** 1.38 (0.96) 0.20 (1.32) 0.80 (0.39)** 0.001 (0.002) 0.004 (0.001)*** 0.05 (0.03) 0.63 (0.03)*** 0.23 (0.57) 0.36 (0.08)** 0.04 (0.09) 0.38 (0.05)*** 0.13 (0.07)* 0.05 (0.02)** 0.0001 (0.0001) 0.00001 (0.00006) 0.0007 (0.002) 0.20(0.03)*** 0.13 (0.01)*** 0.01 (0.002)*** 0.88 (0.22)*** 1.61 (0.34)*** 0.62 (0.03)*** 0.32 (0.53) 0.33 (0.08)*** 0.04 (0.08) 0.40 (0.05)*** 0.14 (0.07)** 0.05 (0.02)** 0.0002 (0.0001)** 0.04 (0.01)*** 0.001 (0.002) 0.16 (0.02)*** 0.12 (0.01)*** 0.01 (0.002)*** 0.58 (0.22)*** 1.58 (0.33)** All banks (19901998) All banks (19992008) All banks (19902008) All banks (19902008) All banks (19901998) 0.45 (0.04)*** 0.29 (0.59) 0.07 (0.05) 0.19 (0.11)* 0.62 (0.06)*** 0.00 (0.04) 0.03 (0.03) 0.0004 (0.0003) 0.0002 (0.001)* 0.0004 (0.0001) 0.23 (0.02)*** 0.02 (0.04) 0.64 (0.53) 0.76 (0.78) 72 9308.43*** 40.26*** All banks (19992008) 0.54 (0.04)*** 0.25 (0.65) 0.20 (0.10)** 0.00 (0.11) 0.40 (0.06)*** 0.07 (0.11) 0.06 (0.04) 0.00002 (0.0001) 0.00002 (0.00006) 0.0003 (0.0004) 0.13 (0.02)*** 0.01 (0.00)*** 0.21 (0.31) 1.31 (0.41)*** 260 564.85*** 254.09***

Interest spread

All banks (Period)

All banks (19902008)

Lagged interest spread Overhead Capital ratio NPL Bank size Non-interest income MSD HHI QI GDP growth Ination FLI LRR Bank rate Tax Constant

0.42 (0.06)*** 15.92 (6.63)*** 1.27 (1.05) 2.18 (1.19)* 0.77(0.71) 0.22(0.85) 0.76 (0.31)** 0.001 (0.001) 0.002 (0.001)*** 0.02(0.02) 0.54 (0.30)* 0.73 (0.26)*** 0.02 (0.03) 2.52 (2.33) 8.70 (3.61)***

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408

N Wald (2 test Sargan test ((2 value)

300 176.23*** 190.75***

Notes: Standard errors are in parentheses. See notes in Table 3 for denitions of variables: interest spread, overhead, capital ratio, NPL, bank size, non-interest income, MSD, HHI, QI, LRR, and Tax. 3. The net interest margin is dened as the difference between total interest and commission received over total earning assets and total interest paid minus fees over total interest bearing liabilities. The impact of nancial liberalization is captured by a nancial liberalization dummy (FLI) (1 for year 1999 and onward, 0 otherwise) as the sector was fully liberalized in 1999. Source: Authors estimation. * Represent signicance at 10% level. ** Indicate signicance at 5 percent level, respectively. *** Indicate signicance at 1 percent level, respectively.

403

Author's personal copy

404 M. Hossain / Journal of Asian Economics 23 (2012) 395408

percent and 63 percent, respectively. This high persistency effect indicates that a major part of spread in Bangladesh is attributed to inefciencies of bank management, particularly in credit allocations and risk management practices. This persistency effect points to the weaknesses of corporate governance of the banking sector as board directors, who are usually nominated by the owners, often dictate managers to maintain a certain level of prot. From operational perspective, the persistency of spread may be linked to the Lazy bank hypothesis (Manove, Padilla, & Pagano, 2001), which argues that the presence of a high level of guarantees (collateral) weakens the banks incentive to evaluate the protability of a planned investment project. High level of spread can also act as a guarantee as it reduces screening and monitoring activities of banks. Overhead cost and NPL are found to be positively and signicantly associated with interest spread for the whole sample period as well as for the post-liberalization period, but not with margin. Both the factors point to management inefciency for which the cost has to be borne by the customers. For the pre-liberalization period (before 1999), capital ratio is found to be negatively associated with spreads. This is consistent with the argument that due to limited channels for raising capital in the 1990s particularly in the absence of a well-functioning capital market, banks were in a strong position to keep the spreads high. The coefcient of bank size is signicant and negatively related to interest margin, indicating that bigger size can signicantly reduces margins. On the other hand, before 1999, bank size was signicant and positively related to spread. This may be due to the fact that larger banks, particularly SCBs maintained a higher spread in the 1990s because of their market power. Market share of deposits (MSD) is found to be negative and signicant, lending support to the small nancial system view of spreads or margins in Bangladesh. This nding calls for rationalization of non-bank savings rates, particularly the NSD certicate (savings) rates and postal savings rates in order to make the market more competitive. Financial liberalization, represented by a dummy (FLI), has negative and marginally signicant impact on spreads, but positive impact on interest margin. Although the FLI dummy is a rough indicator of nancial liberalization, the nding is broadly consistent with our overall ndings that liberalization did not contribute much to the rationalization of spreads. The Herndahl-Hirschman index (HHI) representing market concentration on loans has been included in the model. The HHI is found to be signicant to spreads for all banks for the whole period, 19902008, indicating a less competitive banking sector in Bangladesh. From macroeconomic point of view, quantum index (QI) of production is found to be positive and signicant to spreads. This indicates the existence of a segmented credit market as the QI represents investment behavior of rms. On the other hand, ination is found to be positively associated with spread in the liberalized period. This positive association can be explained by the fact that an increase in ination drives down the real rate of returnthus the adverse effects of ination are compensated with higher spreads. Among the monetary policy variables, only liquidity reserve requirement (LRR), which is currently 20 percent, is found to be negative and signicant to spread. A rise of reserve requirement, which is a kind of nancial taxation on commercial banks, compel banks to increase deposit rates to attract more liquid funds, which eventually decreases the spread. Thus, the LRR could

Table 5 Determinants of interest spreads and margins for private commercial banks (PCBs). PCBs (Period) Interest spread PCBs (19902008) Lagged interest spread Overhead Capital ratio NPL Bank size Other income MSD HHI QI Ination FLI Tax LRR Bank Rate Constant N Wald (2 test Sargan test ((2 value) 0.41 (0.07)*** 16.08 (7.58)* 0.85 (1.16) 2.49 (1.65) 1.17 (0.98) 0.42 (0.90) 1.76 (1.19) 0.02 (0.02) 0.004 (0.001)*** 0.06 (0.03)** 0.24 (0.31) 3.03 (2.91) 0.88 (0.37)*** 0.03 (0.04) 7.76 (4.58)* 235 173.5*** 173.03*** PCBs (19992008) 0.30 (0.09)*** 16.75 (9.60)* 1.39 (2.17) 2.94 (1.83) 1.09 (1.27) 0.40 (1.45) 2.69 (1.46)** 0.01 (0.03) 0.006 (0.001)*** 0.10 (0.05)*** 5.21 (3.67) 0.81 (0.44)** 0.02 (0.04) 8.13 (6.55) 191 128.79*** 121.77** PCBs (old) (19992008) 0.46 3.52 0.39 1.91 0.44 0.63 2.68 0.04 0.01 0.06 0.24 8.27 0.90 0.01 12.63 (0.07)*** (8.20) (1.07) (1.43) (1.24) (0.82) (1.26)** (0.02)** (0.001)*** (0.03)*** (0.34) (3.67)*** (0.35)** (0.03) (5.51)*** Interest margin PCBs (19902008) 0.06 (0.03)** 0.32 (0.25) 0.14 (0.04)*** 0.00 (0.04) 0.54 (0.03)*** 0.05 (0.03)* 0.13 (0.04)*** 0.001 (0.00006)** 0.00002 (0.00002) 0.0007 (0.001) 0.08 (0.01)*** 0.21 (0.01)*** 0.01 (0.00)*** 0.51 (0.11)*** 1.43 (0.17)*** 263 .75*** 229.06*** PCBs (19992008) 0.08 (0.03)*** 0.32 (0.31) 0.00 (0.05) 0.07 (0.05) 0.60 (0.04)*** 0.11 (0.05)*** 0.08 (0.76) 0.001 (0.00007)** 0.000008 (0.00003) 0.001 (0.002) 0.24 (0.02)*** 0.01 (0.00)*** 0.07 0.17 1.41 (0.20)*** 212 473.65*** 196.95*** PCBs (old) (19992008) 0.70 (0.04)*** 1.37 (0.77)* 0.09 (0.11) 0.07 (0.11) 0.18 (0.07)*** 0.07 (0.11) 0.23 (0.12)*** 0.00001 (0.0001) 0.00001 (0.00008) 0.001 (0.004) 0.06 (0.02)*** 0.05 (0.02)** 0.39 (0.34) 0.85 (0.45)** 204 534.04*** 183.23***

185 128.79** 121.78**

Notes: See notes in Tables 3 and 4 for denitions of variables. Source: Authors estimation. * Represent signicance 10% level. ** Represent signicance at 5% level. *** Represent signicance at 1% level.

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408 405

be an effective monetary policy instrument to reduce the spread. However, other policy instruments, such as discount rate and corporate tax rate are insignicant to spread although they have some negative effects on interest rate margins. 5.2.1. Results for the PCBs The determinants of interest rate spreads and margins are analyzed separately for PCBs in Table 5. Lagged spreads and margins are found to be signicant indicating persistency in spreads and margins in the case of PCBs. The coefcient of overhead cost is positive and signicant to spread in the case of all PCBs, but not signicant when only old PCBs (established before 1999) are considered. This implies that high overhead cost leads to high spread particularly in the case of new PCBs that were established after 1999. The Herndahl-Hirschman index is signicantly associated with spreads in the case of old PCBs. When all types of PCBs are considered, it is found signicant only to interest margins. Therefore, it may be concluded from the analysis that among the PCBs, the older PCBs have a certain degree of market power that allows them to keep higher spreads, particularly after 1999. Small share of deposits also contributes to high spreads charged by the old PCBs, that is, segmented deposit market is also partly responsible for higher spreads. The capital ratio is positive and signicant to interest margins for the whole period (19902008), but insignicant for the post-liberalization period indicating that high margins contribute to high bank earnings, which are channeled into the capital base of the PCBs before liberalization. The bank size is negative and signicant in explaining interest margins for all private commercial banks. Financial liberalization (FLI) dummy is not signicant to spreads, but it seems to have caused interest margins of PCBs to widen. Among the monetary policy variables, LRR is negative and signicant to spreads and margins, while the bank rate is negatively and signicantly associated only with the interest margin. The QI is signicant to spreads indicating that segmented credit market has added extra leverage that contributed to high spreads of the PCBs. Thus, from the analysis it has become clearer that the PCBs are the dominant players in the banking sector of Bangladesh, particularly in determining spreads in the post-liberalization period. On the other hand, the newly established PCBs are in a vulnerable situation. Therefore, market power and segmented credit and deposit markets are mainly responsible for maintaining persistently high spreads by the older PCBs in the post-liberalization period. 5.2.2. Results for the SCBs The determinants of interest rate spreads and margins are separately analyzed for the SCBs in Table 6. Interest rate spreads are not found to be persistent in the case of SCBs. Only overhead cost and non-interest incomes are positive and signicant to spreads in SCBs. These two factors broadly trigger to inefciencies in SCBs, which are attributed to (i) government intervention in loan disbursement as well as in day-to-day management; (ii) high administrative costs due to large number of branches as well as employees, and (iii) poor service quality of these banks. Moreover, spreads in SCBs is partly inuenced by the macroeconomic environment as ination is found to be positive and signicant. To sum up, interest rate spreads in the banking sector of Bangladesh is determined by the PCBs in the post-liberalization period. Market power, small share of deposits and segmented credit markets are largely responsible for persistently high spreads in PCBs. Although nancial liberalization has succeeded in reducing the dominance of SCBs and increasing the role of

Table 6 Determinants of interest spreads and margins for state-owned commercial banks (SCBs). SCBs (Period) Interest spread SCBs (19902008) Lagged interest spread Overhead Capital ratio NPL Bank size Non-interest income MSD HHI QI Ination FLI LRR Tax Constant N Wald (2 test Sargan test ((2 value) 0.22 (0.16) 23.48(12.65)** 5.41(2.66)** 0.35(1.27) 5.54(1.63)*** 4.88(2.77)* 0.29(0.97) 0.001(0.001) 0.001(0.001) 0.09(0.04)*** 0.17(0.29) 1.25(0.65)** 4.43 (4.3) 35.86(6.88)*** 53 55.35*** 40.81 SCBs (19992008) 0.07(0.15) 53.95(18.71)*** 6.12(3.17)** 0.11(2.53) 6.36(4.61) 6.04(2.30)*** 0.83(3.62) 0.002(0.002) 0.001(0.001) 0.06(0.07) 1.69(1.77) 7.47 (9.12) 35.52(16.51)*** 37 35.55*** 31.18 Interest margin SCBs (19902008) 0.36 (0.14) 0.03 (0.14) 0.01 (0.03) 0.01 (0.02) 0.01 (0.04) 0.02 (0.02) 0.01 (0.03) 0.00004 (0.0001)*** 0.000006 (0.00001) 0.0009 (0.0005)** 0.01 (0.01) 0.01 (0.02) 0.09 (0.04)*** 0.02 (0.15) 53 134.23*** 41.36
***

SCBs (19992008) 0.14 (0.16) 0.09 (0.15) 0.03 (0.03) 0.02 (0.02) 0.01 (0.05) 0.01 (0.02) 0.04 (0.04) 0.00004 (0.00002)** 0.0008 (0.001) 0.002 (0.002) 0.01 (0.02) 0.08 (0.09) 0.02 (0.17) 37 80.3*** 28.32

Notes: see notes in Tables 3 and 4 for denitions of variables. Source: Authors estimation. * Represent signicance at 10% level. ** Represent signicance at 5% level. *** Represent signicance at 1% level.

Author's personal copy

406 M. Hossain / Journal of Asian Economics 23 (2012) 395408

the private banks in the nancial system, still the banking sector lacks adequate competition and efciency. This indicates that the nancial system lacks proper legal and institutional infrastructure. It is therefore important to develop necessary legal and nancial institutions in order to get the maximum benet of nancial reforms, particularly to narrow the spread (Beck & Levine, 2004; Chinn & Ito, 2006). Development of the capital market and bond market, development of prudential regulatory framework and gradual privatization of SCBs are some of the measures that can be adopted to make the banking sector more competitive in a liberalized environment. In addition to these, rationalization of non-bank rates can help the sector to be competitive. 6. Conclusion This study has attempted to explain why interest spreads are persistently high in Bangladesh and why different banks charge different interest spreads despite major nancial sector reforms undertaken in the 1990s. The results provide support to all four underlying views of the determinants of spreads, namely market structure, small nancial system, risk-based and macroeconomics-based views. This study for the rst time captures the persistency in spreads and margins that are linked to inefciencies of management arising from revealed preferences, lack of risk management practices and technological skills of management. Persistency effect therefore points to the weaknesses of corporate governance as well as lazy bank argument. More specically, high operating costs and the non-performing loan ratio, market power, and segmented credit and deposit markets are responsible for high spreads in Bangladesh. The nancial liberalization commenced in the 1990s increased nancial deepening by ensuring higher levels of savings and investments in Bangladesh. As a result of reforms and liberalization, market power has shifted to private banks from state-owned banks after liberalization. Although nancial liberalization has increased the depth of the nancial sector, it has not succeeded in generating enough competition and efciency in the nancial sector. This suggests that there are pitfalls in institutional (both legal and nancial) development. Thus, further efforts in developing legal and nancial institutions will be necessary to make the sector competitive and reduce spread. Acknowledgments The author gratefully acknowledges the grant received from the BIDS Policy Resource Program (PRP) for this study. The author thanks K.A.S. Murshid for approving this study under the PRP and providing useful inputs to the study. The author also thanks Quazi Kholiquzzaman, Saleh Uddin Ahmed, M.K. Mujeri and other participants for their helpful comments in a seminar organized by the PRP on this paper. The author thanks an anonymous referee and the Editor, M.G. Plummer for useful comments on the paper. Special thanks to Sifat Adiya for providing editing support. The usual disclaimer applies.

Appendix See Tables A.1 and A.2.

Table A.1 An international comparison of the determinants of interest rate spreads and margins. References A. African countries Beck and Hesse (2009) Country/Sample period Uganda. 19992005 Methodology Pooled OLS; median least square; xed effect Determinants of spreads/margins Small market, high operating cost, high ination, high T-bill rate, exchange rate appreciation Low ination, greater number of banks, greater public ownership of banks, poor governance, higher reserve ratio

Crowley (2007)

Average spread: 18%. 18 African countries (Botswana, Ethiopia, Gambia, Ghana, Kenya, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Nigeria, Rwanda, Sierra Leone, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe). 19752004. Average spread: 7% Malawi. 19891999

Cross-section OLS

Chirwa and Mlachila (2004)

Fixed effect, random effect panel regression

Monopoly power, reserve requirements, discount rate, ination

Author's personal copy

M. Hossain / Journal of Asian Economics 23 (2012) 395408 Table A.1 (Continued ) References B. Eastern Caribbean countries Randall (1998) Country/Sample period Average spread: 16.75%
a

407

Methodology

Determinants of spreads/margins

Eastern Caribbean countries (Antigua and Barbuda, Dominica, Grenada, St. Kitts and Nevis, St. Lucia, and St.

Two-stage least square estimation

Operating cost, loan loss provision and reserve costs accounts for 75% of the observed interest spread. Among policy variables, statutory deposit requirements, strict loan loss provision, and bank xed costs are important for high spread.

Vincent and the Grenadines). 19911996 Average spread: 7.3% C. Latin American countries Brock and Rojas-Suarez (2000)

Latin America. 19911995 Argentina (12.9%), Bolivia (7.1%), Colombia (21%), Chile (11.6%), Peru (20%), Mexico (7.7%). Colombia. 19741996 Average spread: 1632% (19741988); 2519% (19881996)a

Two-step regression

Capital ratio, cost ratio, liquidity ratio, interest rate volatility, ination

Barajas et al. (1999)

Two-stage least squares

Operating cost, nancial taxation, loan quality and market power

D. OECD countries Saundars and Schumacher (2000) Angbazo (1997)

7 OECD countries (Germany, Spain, France, UK, Italy, USA, Switzerland) US 19891993

Two-step regression

Capital ratio, monopoly power, volatility of interest rates De fault risk, opportunity cost of non-interest bearing reserves, leverage, management efciency NPL, high administrative cost and non-competition Limited competition, overstafng, high administrative costs, NPLs. Inelasticity of deposit, liquidity, NPL Bank size, real T-bill rate, liquidity ratio, concentration, ination, GDP growth, institutional deciencies, overhead cost Ratio of equity to lagged total asset, ratio of loans to total assets, foreign ownership, bank size, overhead cost, ination rate, short-term market interest rate

E. South Asian countries Mujeri and Islam (2008)

Bangladesh: 20012007 Average spread: 6% Bangladesh: 20032006 Average spread: 5.6% Pakistan. 19982005 Cross-country. 86 countries; 20002004; Average spread: 5% Cross-country (80). 19881995

Summary statistics and conceptual argument Summary statistics

Ahmed and Islam (2006)

Khawaja and Din (2007) F. Cross-country analysis Beck and Hesse (2009)

Fixed effect model Cross-sectional OLS

Demirguc-Kunt and Huizinga (1999)

Cross-sectional OLS

Estimation of spread depends on a particular denition.

Table A.2 List of banks considered for the analysis. State-owned commercial banks (SCBs) Agrani Bank Ltd Janata Bank Ltd Sonali Bank Ltd Rupali Bank Ltd Specialized banks (SBs) BASIC Bank Limited Bangladesh Krishi Bank Bangladesh Shilpa Bank Private commercial banks (PCBs) AB Bank Limited A L-Arafa Islami Bank Ltd BRAC Bank Limited Bangladesh Commerce Bank Ltd Bank Al-Falah Limited Bank Asia Dhaka Bank Dutch-Bangla Bank Ltd EXIM Bank Limited Eastern Bank Limited First Security Islami Bank Ltd ICB Islami Bank IFIC Bank Limited Islami Bank Bangladesh Limited Mercantile Bank Ltd Foreign commercial banks (FCBs) Citibank N.A Commercial Bank of Ceylon Habib Bank Ltd Standard Chartered Bank State Bank of India The Hong Kong and Sanghai Bank Ltd Woori Bank

Author's personal copy

408 Table A.2 (Continued ) State-owned commercial banks (SCBs) Specialized banks (SBs) Private commercial banks (PCBs) Mutual Trust Bank National Bank Limited National Credit and Commerce Bank Limited One Bank Limited Premier Bank Limited Prime Bank Ltd Pubali Bank Ltd Rupali Bank Ltd Shahjalal Bank Ltd Southeast Bank Ltd Standard Bank Ltd The City Bank Ltd Trust Bank United Commercial Bank Uttara Bank Limited Foreign commercial banks (FCBs) M. Hossain / Journal of Asian Economics 23 (2012) 395408

References
Ahmed, S., & Islam, M. E. (2006). Interest rate spread in bangladesh: An analytical review. Dhaka: Policy Note Series PN 0701, Policy Analysis Unit, Bangladesh Bank. Angbazo, L. (1997). Commercial bank net interest margins, default risk, interest-rate risk, and off-balance sheet banking. Journal of Banking and Finance, 21, 5587. Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68(1), 2951. Arellano, M., & Bond, S. (1991). Some tests of specication for panel data: Monte carlo evidence and an application to employment equations. Review of Economic Studies, 58(2), 277297. Baltagi, & Badi, H. (2005). Econometric analysis of panel data. West Sussex: John Wiley and Sons. Barajas, A., Steiner, R., & Salazar N. (1999). Interest spreads in banking in colombia, 19741996, IMF Staff Papers, 46(2). Beck, T., & Hesse, H. (2009). Why are interest spreads so high in Uganda? Journal of Development Economics, 88, 192204. Beck, T., & Levine, R. (2004). Legal institutions and nancial development. In M. Claude & M. Shirley (Eds.), Handbook of new institutional economics. The Netherlands: Kluwer Dordrecht. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87, 115143. Brock, P. L., & Rojas-Suarez, L. (2000). Understanding the behavior of bank spreads in Latin America. Journal of Development Economics, 63, 113134. Chinn, M. D., & Ito, H. (2006). What matters for nancial development? Capital controls, institutions and interactions. Journal of Development Economics, 81(1), 163192. Chirwa, E.W., & Mlachila, M. (2004). Financial reforms and interest rate spreads in the commercial banking system in Malawi. IMF Staff Papers, 51(1). Crowley, J. (2007). Interest rate spreads in English-speaking African countries. IMF Working Paper WP/07/101, International Monetary Fund, Washington DC. Demirguc-Kunt, A., & Huizinga, H. (1999). Determinants of commercial bank interest margins and protability: Some international evidence. World Bank Economic Review, 13, 379408. Denizer, C. (1999). Foreign entry in Turkeys banking sector, 19801997. (unpublished; Washington, IFC/World Bank). Honohan, P. (1999). Consequences for Greece and Portugal of the opening-up of the European banking market (unpublished; Washington: Development Economics Group, World Bank). JudsonF R.A., & OwenF A.L., (1999). Estimating dynamic panel data models: A guide for macroeconomists. Economic Letters, 65, 915. Khan, M.S., & Senhadji, A.S. (2000). Financial Development and Economic Growth: An Overview, IMF Working Paper 209 (Washington: International Monetary Fund). Khawaja, I., & Din, M. (2007). Determinants of interest spread in Pakistan. PIDE Working Papers, 2007:22. King, R., & Levine, R. (1993). Finance and growth: Schumpeter might be right. Quarterly Journal of Economics, 108(3), 717737. Levine, R. (1997). Financial development and economic growth: Views and agenda. Journal of Economic Literature, 35(2), 688726. Manove, M., Padilla, A. J., & Pagano, M. (2001). Collateral versus project screening: A model of lazy banks. Rand Journal of Economics, 32(4), 726744. McKinnon, R. I. (1973). Money and capital in economic development. Washington: The Brookings Institution. Mujeri, M.K., & Islam, M.E. (2008). Rationalizing interest rate spread in the banking sector: some policy suggestions, Policy Paper No. PP 0804, Policy Analysis Unit, Bangladesh Bank, Dhaka. Mujeri, M. K., & Younus, S. (2009). An analysis of interest rate spread in the banking sector in Bangladesh. The Bangladesh Development Studies, XXXII(4), 133. Nickell, S. (1981). Biases in dynamic models with xed effects. Econometrica, 49, 14171426. Randall, R. (1998). Interest rate spreads in the Eastern Caribbean. IMF Working Paper WP/98/59, International Monetary Fund, Washington D.C. Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables. Econometrica, 26(3), 393415. Saundars, A., & Schumacher, L. (2000). The determinants of bank interest rate margins: An international study. Journal of International Money and Finance, 19, 813832. Shaw, E. M. (1973). Financial deepening in economic development. Oxford University Press: Oxford. Wooldridge, J. M. (2002). Econometric analysis of cross sectional and panel data. Cambridge: The MIT Press.